[lxc-devel] lxcfs and print_subsystems()

[Serge Hallyn]

Quoting Dietmar Maurer (dietmar at proxmox.com):
> Whenever I do a 'ps' inside a container, the host logs:
> 
> Nov 18 12:25:20 elsa lxcfs[9376]: hierarchies: 1: name=systemd
> Nov 18 12:25:20 elsa lxcfs[9376]: 2: cpuset
> Nov 18 12:25:20 elsa lxcfs[9376]: 3: cpu,cpuacct
> Nov 18 12:25:20 elsa lxcfs[9376]: 4: blkio
> Nov 18 12:25:20 elsa lxcfs[9376]: 5: memory
> Nov 18 12:25:20 elsa lxcfs[9376]: 6: devices
> Nov 18 12:25:20 elsa lxcfs[9376]: 7: freezer
> Nov 18 12:25:20 elsa lxcfs[9376]: 8: net_cls,net_prio
> Nov 18 12:25:20 elsa lxcfs[9376]: 9: perf_event
> Nov 18 12:25:20 elsa lxcfs[9376]: 10: hugetlb
> 
> Is that really required? Or can we remove that?
> 
> Index: new/cgfs.c
> ===================================================================
> --- new.orig/cgfs.c
> +++ new/cgfs.c
> @@ -208,7 +208,7 @@ static bool collect_subsystems(void)
>                         goto out;
>         }
>  
> -       print_subsystems();
> +       //print_subsystems();
>         bret = true;
>  
>  out:

That should only be done at startup.  If you're getting this at every
'ps' in a container, it suggests that lxcfs is crashing and being
restarted.  What version of lxcfs are you on?  Can you stop lxcfs, start
it by hand inside gdb, and get a stack trace when it dies?

-serge